Author + information
- Received February 19, 2014
- Revision received May 7, 2014
- Accepted May 13, 2014
- Published online August 5, 2014.
- Yao-Jun Zhang, PhD∗,†,
- Javaid Iqbal, MRCP, PhD∗,
- Carlos M. Campos, MD∗,
- David V. Klaveren, MSc‡,
- Christos V. Bourantas, MD∗,
- Keith D. Dawkins, MD§,
- Adrian P. Banning, MD‖,
- Javier Escaned, MD, PhD¶,
- Ton de Vries, MSc#,
- Marie-Angèle Morel, BSc#,
- Vasim Farooq, MD∗,
- Yoshinobu Onuma, MD∗,
- Hector M. Garcia-Garcia, MD, PhD∗,
- Gregg W. Stone, MD∗∗,
- Ewout W. Steyerberg, PhD‡,
- Friedrich W. Mohr, MD†† and
- Patrick W. Serruys, MD, PhD∗,‡‡∗ ()
- ∗Thoraxcenter, Erasmus Medical Center, Rotterdam, the Netherlands
- †Nanjing First Hospital, Nanjing Medical University, Nanjing, China
- ‡Department of Public Health, Erasmus Medical Center, Rotterdam, the Netherlands
- §Boston Scientific Corporation, Natick, Massachusetts
- ‖Oxford University Hospitals, Oxford, United Kingdom
- ¶Clinico San Carlos University Hospital, Madrid, Spain
- #Cardialysis BV, Rotterdam, the Netherlands
- ∗∗Columbia University Medical Center/New York-Presbyterian Hospital, Cardiovascular Research Foundation, New York, New York
- ††Herzzentrum Universität Leipzig, Leipzig, Germany
- ‡‡Imperial College London, London, United Kingdom
- ↵∗Reprint requests and correspondence:
Dr. Patrick W. Serruys, Department of Interventional Cardiology, Thoraxcenter Erasmus Medical Center, ’s Gravendijkwal 230, Ba-583, Rotterdam, Zuid Holland 3015 GD, the Netherlands.
Background The results of SYNTAX trial have been reported based on “corelab” calculated SS (cSS). It has been shown that reproducibility of SS is better among the core laboratory technicians than interventional cardiologists. Thus, the prognostic value and clinical implication of the “site” SYNTAX SS (sSS) remain unknown.
Objectives The study sought to evaluate the prognostic value and clinical implication of the sSS after percutaneous coronary intervention (PCI) or coronary artery bypass graft (CABG) surgery in the randomized SYNTAX trial.
Methods The sSS was calculated by the site investigators before randomization in the SYNTAX trial. New tertiles based on the sSS were defined with low (0 to 19), intermediate (20 to 27), and high (≥28) scores. The clinical endpoints were compared between PCI and CABG by Kaplan-Meier estimates, log-rank comparison, and Cox regression analyses using the new tertiles. The sSS-based SS II was calculated and its predictive performance was evaluated.
Results The mean difference in cSS and sSS is 3.8 ± 11.2, with a mean absolute difference of 8.9 ± 7.8. In the overall cohort, using sSS there was a higher incidence of major adverse cardiac and cerebrovascular events (MACCE) at 5-year follow-up in the PCI group for low (31.9% vs. 24.5%; p = 0.054), intermediate (39.5% vs. 29.5%; p = 0.019), and high (43.0% vs. 31.4%; p = 0.003) tertiles, compared with the CABG group. Similarly, in the 3-vessel disease subgroup, 5-year MACCE rates were higher in PCI group in all tertiles. Conversely, in the left main subgroup, MACCE rates were similar for PCI and CABG groups in all tertiles. The sSS-based SS II (c-index: 0.736) had predictive performance similar to the cSS-based SS II (c-index: 0.744), with net reclassification index of –0.0062 (p = 0.79).
Conclusions Appropriate training and unbiased assessment are needed when using SS in clinical decision making. sSS and tertiles based on sSS showed poor discrimination among low, intermediate, and high-risk groups. However, combining clinical factors with sSS retained the predictive performance of SS II. (SYNTAX Study: TAXUS Drug-Eluting Stent Versus Coronary Artery Bypass Surgery for the Treatment of Narrowed Arteries; NCT00114972)
Interventional cardiologists and surgeons in the SYNTAX (SYNergy Between PCI With TAXUS and Cardiac Surgery) trial originally used the SYNTAX score (SS) to extract objective information from the coronary angiogram on the technical challenges posed by coronary anatomy to percutaneous coronary intervention (PCI) and to facilitate discussions made by the heart team (1). Subsequently, it became apparent that the SS had a prognostic value to predict short- and long-term outcomes (2–4). The European and American revascularization guidelines currently recommend the SS to guide the heart term in decision making (5–7). Moreover, high-risk SS category is a key inclusion/exclusion criterion, imposed by the U.S. Food and Drug Administration, in several ongoing randomized controlled trials, including EXCEL (Evaluation of Xience Prime or Xience V versus Coronary Artery Bypass Surgery for Effectiveness of Left Main Revascularization), PARTNER-II (Placement of AoRtic TraNscathetER Valves), and SURTAVI (Surgical Replacement and Transcatheter Aortic Valve Implantation).
Knowledge of variability in calculating SS is of paramount importance at a time of its widespread use as a clinical decision-making tool. Risk stratification of patients in the SYNTAX trial was based on “corelab” SS (cSS), which was calculated by the core laboratory technicians blinded to the treatment group. However, all sites participating in SYNTAX also had a “site” SS (sSS) calculated by the site investigators, who at that time had no knowledge of the prognostic significance of the SS. The discrepancy between cSS and sSS has been previously highlighted in the SYNTAX trial (8). Recently, Généreux et al. (9) have demonstrated that interventional cardiologists underestimate the number of lesions, bifurcation, and the presence of small-vessel disease, resulting in a lower SS than that reported by the core laboratory technicians. The significance and prognostic value of the sSS in the SYNTAX trial have not been evaluated.
This study aimed to examine the difference between the cSS and sSS in the randomized SYNTAX trial and the prognostic performance of the sSS in assessing outcomes among patients undergoing PCI or coronary artery bypass grafting (CABG). We also investigated the predictive performance of the SYNTAX II score, a recently developed risk score that combines the anatomic SS with clinical variables to predict long-term outcome of PCI and CABG when calculated using either cSS or sSS.
The SYNTAX randomized trial
The SYNTAX trial (NCT00114972) was a prospective, multicenter, randomized trial to investigate subjects with unprotected left main coronary artery (ULMCA) disease (isolated or associated with 1-vessel, 2-vessel, or 3-vessel disease), or de novo 3-vessel disease (3VD) (10). Eligible patients were randomized on a 1:1 ratio to CABG (n = 897) or PCI with Taxus Express paclitaxel-eluting stent (Boston Scientiﬁc Corporation, Natick, Massachusetts; n = 903) and followed up for 5 years. The primary clinical endpoint of the SYNTAX trial was a composite of major adverse cardiac or cerebrovascular events (MACCE) (e.g., death from any cause, stroke, myocardial infarction [MI], or repeat revascularization) at 1-year follow-up. Secondary endpoints included the incidence of MACCE and its components at 1-month, 6-month, 3-year, and 5-year follow-up. An independent clinical event committee comprising interventional cardiologists, cardiac surgeons, and a neurologist adjudicated all events.
The SS calculation
During the local heart team meeting, the cardiac surgeon and interventional cardiologist systematically reviewed the coronary angiogram and specified the number of coronary lesions, along with their angiographic location and characteristics. Diagnostic angiograms were scored according to the SS algorithm (1). Each significant lesion (defined as a diameter stenosis of ≥50% in ≥1.5-mm vessels) is visually assessed and awarded a score related to location and severity of the coronary lesion. Additional points are given for total occlusion, bifurcation or trifurcation lesion, aorto-ostial lesion, severe tortuosity, heavy calcification, thrombus, and diffusely diseased segment. The sSS was calculated during the local heart team meeting before randomization. Calculation of the cSS was done by an independent core laboratory (Cardialysis BV, Rotterdam, the Netherlands), blinded to treatment assignment.
The SS II calculation
The SS II has been previously generated by a combination of anatomic cSS and clinical factors (age, creatinine clearance, left ventricular ejection fraction, presence of ULMCA disease, peripheral vascular disease, sex, and chronic obstructive pulmonary disease) to predict 4-year mortality risk after PCI or CABG (11). Using the same model, we also have calculated SS II using the sSS and using the sSS-based SS II to predict 4-year mortality and make a treatment recommendation.
Continuous variables are expressed as mean ± SD and categorical variables are shown as counts and percentages of the total. Bland-Altman plots were used to compare the cSS and sSS. The predictive values of anatomic cSS and sSS were compared by evaluating differences in the area under the receiver operating characteristic curves (AUC) and standard errors using the Delong method. The agreement between observed and predicted risks for cSS and sSS was assessed with the Hosmer-Lemeshow test. Comparisons of 5-year clinical outcomes between CABG and PCI were conducted with the Kaplan-Meier method and the log-rank tests. By Cox regression analyses, the relative risks were shown as hazard ratios (HRs) and 95% confidence interval (CI). Two patient subsets were predefined in this study: patients with ULMCA (with or without additional vessel involvement), and those with 3VD in the absence of left main coronary disease. To calculate the SS II, multiple imputations (5×) of missing values was performed with an advanced imputation strategy, which takes into account the correlation between all potential predictors (11,12). The performance of the sSS-based SS II was evaluated using c-statistics (Harrell’s c-index) (13), calibration plots (14), reclassification table, and net reclassification index (NRI) (15,16). A probability value of less than 0.05 was considered statistically significant. All analyses were undertaken using SPSS 20.0 (IBM Corporation, Armonk, New York).
SS by the core laboratory and site
The cSS was calculated in 99.4% (n = 1,789) of patients, missed in 0.6% (n = 11) of patients. The sSS was available in 100% (n = 1,800). The cSS and sSS distributions are shown in Figure 1. The mean SS was 28.7 ± 11.4 (range: 0 to 83) for the corelab and 24.9 ± 10.2 (range: 3 to 88) for the site. The mean difference between cSS and sSS is 3.8 ± 11.2, with an absolute difference of 8.9 ± 7.8 (Central Illustration). The sSS was numerically identical to the cSS in 99 patients (5.5%), underestimated in 1,106 (61.8%), and overestimated in 584 (32.7%). A significant correlation (r = 0.49; p < 0.05) was found between the cSS and sSS as shown in Figure 2A. However, Bland-Altman plots showed that the limits of agreement (2 SDs that describe the range for 95% comparison points) were very wide (–18.57 to +26.23 score), indicating a poor agreement between the corelab and site calculated SS (Fig. 2B). Only 882 (49.3%) patients have concordant scores according to the corelab SS-defined tertiles (Table 1).
Clinical endpoints using sSS and established SS tertiles
Using the sSS, patients were grouped into traditional SYNTAX tertiles of low (≤22 score, n = 823), intermediate (23 to 32 score, n = 619), and high (≥33 score, n = 358) scores (Central Illustration).
In the overall cohort, using the sSS, MACCE rates in the low and intermediate tertiles were significantly higher in the PCI group than in the CABG group (p = 0.016, p = 0.002, respectively); MACCE rate in the high-score tertile was numerically but not statistically higher with PCI compared with CABG (42.0% vs. 33.8%; p = 0.105) (Central Illustration). In the ULMCA subgroup, the 2 revascularization strategies had similar MACCE rates during the 5-year follow-up period (Fig. 3A). In the 3VD subgroup, the incidence of MACCE was significantly higher in the PCI group than in the CABG group in all tertiles (low, 35.3% vs. 25.9%; p = 0.024; intermediate, 40.1% vs. 26.4%; p = 0.005; high, 46.1% vs. 24.3%; p = 0.006) (Fig. 3B).
Clinical endpoints using sSS and new tertiles based on sSS
The actual tertiles according to the sSS were low (≤19 score, n = 616), intermediate (20 to 27 score, n = 567), and high (≥28 score, n = 617).
In the overall cohort, there was a trend toward higher incidence of MACCE in the PCI group for low scores, compared with the CABG group, but this difference was not statistically significant (31.9% vs. 24.5%; p = 0.054) (Fig. 4A). Risk of MACCE for intermediate and high scores was significantly increased with PCI versus CABG (39.5% vs. 29.5%; p = 0.019; 43.0% vs. 31.4%; p = 0.003, respectively) (Fig. 4A).
In the ULCMA disease subgroup (n = 705), MACCE rates were comparable and did not differ significantly between the 2 groups in each tertile (Fig. 4B). In the 3VD subgroup (n = 1,095), all groups of low, intermediate, and high scores showed a significantly higher incidence of MACCE in the PCI group than those in the CABG group at 5-year follow-up (Fig. 4C).
There was no significant difference in death, stroke, and repeat revascularization between PCI and CABG in the low tertiles, but a higher risk of MI in the PCI group (p = 0.002) (Fig. 5). In the intermediate tertiles, a significantly higher proportion of patients had repeat revascularization after PCI than after CABG (HR, 2.65; 95% CI: 1.72 to 4.08; p < 0.001). In the high tertiles, there was a significantly higher risk of death, MI, and repeat revascularization in the PCI group, but an insignificantly lower risk of stroke (p = 0.177) (Fig. 5).
Substantial changes in treatment decision using sSS
The predictive accuracy for 4-year mortality was modest with cSS (AUC, 0.57; 95% CI: 0.54 to 0.59) and dropped further with sSS (AUC, 0.55; 95% CI: 0.53 to 0.57). There was a poor agreement between predicted and observed MACCE for the sSS and cSS, using Hosmer-Lemeshow test (p = 0.51, p = 0.31, respectively). Furthermore, the treatment decision based on SS tertile changed in more than one-third of patients depending on whether cSS or sSS is used (Table 2). For 9.4% of patients, the treatment decision changed from PCI or CABG to CABG and for 25.7% the treatment recommendation changed from CABG to PCI or CABG.
Minimal changes in treatment decision using sSS-based SS II
The sSS-based SS II had predictive power similar to the cSS-based SS II (Harrell’s c-index: 0.736, 0.744, respectively). The calibration plots of the sSS-based SS II showed a good agreement between the observed and predicted risk of mortality (Fig. 6). The recommendations according to the original cSS- and the sSS-based SS II models did not warrant a change in revascularization strategy (from PCI to CABG or CABG to PCI) in a large majority of patients (Table 3). A large majority (83%) of patients had the same recommendation: only eligible for CABG (n = 346; 19.2%); only eligible for PCI (n = 76; 4.2%); or potentially amenable to both types of revascularization (n = 1,070; 59.4%). Patients with or without events are reclassified by the sSS-based SS II model, and their results are presented in Table 4 (NRI, –0.0062; p = 0.79).
This post-hoc study of the SYNTAX randomized trial has highlighted a significant difference in the anatomic SS calculated by the site and corelab and the calculation of tertiles based on the sSS, which compromised the ability of the SS to distinguish the low-, intermediate-, and high-risk patients, especially in the subgroup of 3VD. However, the prognostic performance of SS II, which combines clinical variables with the SS, remained largely unaffected whether the cSS or the sSS was used.
Differences in site and corelab SS
There is a significant absolute difference in the SS calculated by the site investigators and the core laboratory analysts in the SYNTAX trial. Lack of advanced training at each site is a plausible underlying reason for this difference (9,17). Appropriate training can significantly reduce intraobserver and interobserver variability of the SS calculation by interventional cardiologists (9). A study comparing SS calculation by interventional cardiologists and angiographic core laboratory technicians has shown that interobserver agreement was initially poor among interventional cardiologists (κ = 0.33), but improved substantially after advanced training (κ = 0.76).
Although SS is a continuous variable, in practice it is used as a categorical variable, with 3 categories determined by cSS cutoff values of 22 and 33. These SS categories are being extensively used in clinical practice and the guidelines to decide revascularization strategy and in the ongoing randomized trials (e.g., EXCEL, PARTNER-II, and SURTAVI) as inclusion/exclusion criteria (18). However, our data highlight the variability in calculating SS, which may have a significant impact on decision making and potential consequences on patient outcomes based on the choice of revascularization strategy.
Outcomes based on sSS tertiles
Using the new tertiles (based on sSS) in the ULMCA disease subgroup, there were no statistical differences in MACCE rates between CABG and PCI in all tertiles. These findings are consistent with studies demonstrating comparable clinical outcomes for patients with left main disease undergoing either PCI or CABG, especially in low-risk patients (19,20). Park et al. (19) have reported that PCI with sirolimus-eluting stents was noninferior to CABG with respect to MACCE at 2 years in the treatment of patients with unprotected left main coronary artery stenosis. Because the newer-generation drug-eluting stents have been shown to significantly reduce the stent-related adverse events (21,22), we speculate that the ongoing EXCEL trial will show noninferiority of PCI for treating the majority of patients with ULMCA disease.
It is noteworthy that for new site-based tertiles, not only was a comparable MACCE in the low tertile but also the lowest incidence of MACCE was in the intermediate tertile for both CABG and PCI groups. This is somewhat similar to the findings in the FREEDOM (Future Revascularization Evaluation in Patients with Diabetes Mellitus: Optimal Management of Multivessel Disease) trial, which suggested no differential treatment effect according to the category of the SS (≤22 and ≥33) in patients with diabetes and complex coronary artery disease (23). Indeed, SS as an anatomic tool without taking into consideration the clinical risk profile of an individual patient is suboptimal in predicting all the clinical outcomes. Therefore, attempts have been made to combine anatomic SS with clinical factors (e.g., Logistic Clinical SS) to accurately predict individual patients’ risk (24).
sSS–based SS II
Recently proposed SS II, using anatomic and clinical variables, has been shown to significantly improve decision making between CABG and PCI compared with using anatomic SS alone (11). In the present study, although the site investigators frequently underestimated the sSS and using it changed treatment decisions, its influence on the overall capability of the SS II to predict prognosis and guide decision making was limited. This suggests that clinical variables were more powerful and counteracted any variability in calculating SS. It is plausible that this superiority of SS II stems from the fact that it used SS as a continuous variable and not as a categorical variable. This may circumvent the limitation, outlined above, of the dependence of classification agreement in categorical variables on variability in repeated measurements and frequency distribution of the studied population.
Clinical implication and future perspective
This study has many potential implications on clinical practice. Revascularization guidelines strongly recommend that heart teams use SS for decision making. However, it is essential to highlight that adequate training of the staff calculating SS is vital. It also may be possible in the near future to develop automated algorithms and software to calculate SS from coronary angiography or noninvasive multislice computed tomography (25). Furthermore, it is noteworthy that the uptake of anatomic SS in clinical practice has been modest, despite recommendation by the guidelines. It is probably attributable to variability in calculating SS and lack of clinical risk factors, which physicians generally believe to be equal or more important. Finally, it is appropriate to suggest that the heart teams use SS II, instead of anatomic SS, for decision making to refer the patients for either surgery or PCI.
This is a post-hoc study from the SYNTAX randomized trial. At the time of the trial, an online calculator for SS was not available, and the trial sites may have had limited or varying experience in calculating SS, which may have potential influence on the observed variability.
The anatomic SS was frequently underestimated by the site. Advanced training and unbiased assessment is mandatory in clinical use of the SS. The sSS and tertiles based on sSS showed poor discrimination between low-, intermediate-, and high-risk groups. However, combining clinical factors with sSS retained the predictive performance of SS II.
COMPETENCY IN MEDICAL KNOWLEDGE: Training and unbiased assessment are needed to apply the coronary anatomic SYNTAX score in clinical decision making. Combining clinical factors with the SYNTAX score (referred to as the SYNTAX-II score) reduces variability.
TRANSLATIONAL OUTLOOK: The ongoing SYNTAX-II study will prospectively evaluate the predictive value of the SYNTAX-II score with regard to both short- and long-term clinical outcomes as a guide to referral of patients with coronary artery disease for either surgical or percutaneous revascularization.
The authors express their gratitude to all of the study centers and participants in the SYNTAX trial, who made this study possible.
The SYNTAX trial was funded by Boston Scientiﬁc. Dr. Banning has received speaker and advisory honoraria from Medtronic, Boston Scientific, and Abbott Vascular. Dr. Dawkins is a full-time employee and holds stock in Boston Scientiﬁc. Dr. Stone has served as consultant for Boston Scientific. All other authors have reported that they have no relationships relevant to the contents of this paper to disclose.
- Abbreviations and Acronyms
- area under the receiver operating characteristic curves
- coronary artery bypass grafting
- confidence interval
- ‘corelab’ SYNTAX score
- hazard ratio
- myocardial infarction
- major adverse cardiac or cerebrovascular event(s)
- net reclassification index
- percutaneous coronary intervention
- SYNTAX score
- ‘site’ SYNTAX score
- SYNergy Between PCI With TAXUS and Cardiac Surgery
- unprotected left main coronary artery
- 3-vessel disease
- Received February 19, 2014.
- Revision received May 7, 2014.
- Accepted May 13, 2014.
- American College of Cardiology Foundation
- Garg S.,
- Serruys P.W.,
- Silber S.,
- et al.
- Wykrzykowska J.J.,
- Garg S.,
- Girasis C.,
- et al.
- Valgimigli M.,
- Serruys P.W.,
- Tsuchida K.,
- et al.,
- ARTS II
- Task Force Members,
- Montalescot G.,
- Sechtem U.,
- Achenbach S.,
- et al.
- Fihn S.D.,
- Gardin J.M.,
- Abrams J.,
- et al.,
- American College of Cardiology Foundation; American Heart Association Task Force on Practice Guidelines; American College of Physicians; American Association for Thoracic Surgery; Preventive Cardiovascular Nurses Association; Society for Cardiovascular Angiography and Interventions; Society of Thoracic Surgeons
- Fihn S.D.,
- Gardin J.M.,
- Abrams J.,
- et al.,
- American College of Cardiology Foundation/American Heart Association Task Force
- Généreux P.,
- Palmerini T.,
- Caixeta A.,
- et al.
- Farooq V.,
- van Klaveren D.,
- Steyerberg E.W.,
- et al.
- Steyerberg E.W.
- Farooq V,
- Head SJ,
- Kappetein AP,
- Serruys PW
- Shiomi H.,
- Morimoto T.,
- Hayano M.,
- et al.,
- CREDO-Kyoto PCI/CABG Registry Cohort-2 Investigators
- Valenti R.,
- Migliorini A.,
- Parodi G.,
- et al.
- Farooq V.,
- Vergouwe Y.,
- Räber L.,
- et al.
- Papadopoulou S.L.,
- Girasis C.,
- Dharampal A.,
- et al.